FWIW Seeing the same thing with Opus 5. Not sure if the models are being over optimized for agentic tasks or over reliance on synthetic data, because since Opus

FWIW Seeing the same thing with Opus 5. Not sure if the models are being over optimized for agentic tasks or over reliance on synthetic data, because since Opus 4.6 I feel like emotional/conversational intelligence has been on a decline (Fable being the exception)